ARTICLE

Received 22 May 2015 | Accepted 5 Apr 2016 | Published 10 May 2016

DOI: 10.1038/ncomms11503

OPEN

Atomistic understanding of cation exchange in PbS

nanocrystals using simulations with pseudoligands
Zhaochuan Fan1, Li-Chiang Lin1, Wim Buijs1, Thijs J.H. Vlugt1 & Marijn A. van Huis2

Cation exchange is a powerful tool for the synthesis of nanostructures such as coreshell
nanocrystals, however, the underlying mechanism is poorly understood. Interactions of
cations with ligands and solvent molecules are systematically ignored in simulations. Here,
we introduce the concept of pseudoligands to incorporate cation-ligand-solvent interactions
in molecular dynamics. This leads to excellent agreement with experimental data on cation
exchange of PbS nanocrystals, whereby Pb ions are partially replaced by Cd ions from
solution. The temperature and the ligand-type control the exchange rate and equilibrium
composition of cations in the nanocrystal. Our simulations reveal that Pb ions are kicked out
by exchanged Cd interstitials and migrate through interstitial sites, aided by local relaxations
at coreshell interfaces and point defects. We also predict that high-pressure conditions
facilitate strongly enhanced cation exchange reactions at elevated temperatures. Our
approach is easily extendable to other semiconductor compounds and to other families of
nanocrystals.

1 Process and Energy Department, Delft University of Technology, Leeghwaterstraat 39, 2628CB Delft, The Netherlands. 2 Soft Condensed Matter, Debye
Institute for Nanomaterials Science, Utrecht University, Princetonplein 5, 3584CC Utrecht, The Netherlands. Correspondence and requests for materials
should be addressed to M.A.v.H. (email: M.A.vanHuis@uu.nl).

NATURE COMMUNICATIONS | 7:11503 | DOI: 10.1038/ncomms11503 | www.nature.com/naturecommunications

ARTICLE

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11503

o synthesize novel nanocrystals (NCs) and heteronanocrystals (HNCs) with designed shape and structure, cation
exchange (CE) has recently been extensively explored and
applied111. This process involves colloidal ionic NCs whereby
the cations in the NCs are (partially) replaced by the cations from
the surrounding solution. The CE technique has greatly boosted
the syntheses of nanostructures with various morphologies.
Starting from elementary NCs that are synthesized using wellknown synthesis routes such as hot injection12, the CE technique
can be used to modify nanostructures into other novel
nanostructures or heterostructures such as binary rods13 and
core/shell structures10,14. The CE technique can be extremely
powerful when combined with other advanced (post)syntheses
techniques such as oriented attachment and seeded growth,
which leads to many novel nanostructures including
nanooctapods8, nanorod couples15 and 2D honeycomb
superlattices16.
Pb-Cd exchange (that is, replacing Pb2 by Cd2 ) in lead
chalcogenide NCs is a typical CE process that produces PbE/CdE
(E S, Se, Te) core/shell HNCs14,17,18. This CE process for
colloidal PbS NCs can be expressed as follows:

PbSsolid ! Pb2liquid
CdSsolid
Cd2liquid

A typical experimental route of this CE consists of two steps14,19:

(1) synthesis of the PbS parent NCs, and (2) immersing the
parent NCs in Cd-oleate solutions and applying heat for 0.510 h
at 373473 K. PbS has a rocksalt (RS) crystal structure where the
atoms have a sixfold coordination, whereas CdS has a zinc blende
(ZB) crystal structure where the atoms have a fourfold
coordination. The lattice mismatch between these two
structures is o2% (ref. 20). Pb-Cd CE in PbE NCs has often
been described as a self-limiting process given that a core/shell
structure is generally found in the product. Complete conversion
is usually difcult to achieve18,19,21. Compared to other types of
CE4,22 (for example, Cd-Ag) that take place spontaneously at
ambient temperature and at a shorter timescale, Pb-Cd CE
requires a relatively high temperature and a long time to
overcome a large energy barrier1.
Although the Pb-Cd CE process has been studied by modern
experimental techniques such as energy dispersive X-ray spectrometry, high-resolution transmission electron microscopy
(HRTEM) and 3D electron tomography17,19,21,23,24, a detailed
understanding of the atomistic mechanism is still missing.
Theoretical investigation of this CE process using molecular
simulations also remains a great challenge. Understanding the
thermodynamics and kinetics of CE at the atomic scale is crucial
for the future development of the precise control of this
technique. In this study, classical molecular dynamics (MD)
simulations are used to study Pb-Cd CE in PbS NCs within a
solution. Our results indicate that the rate of CE can be controlled
by changing the ligand type, or by adjusting the temperature and/
or pressure. By changing the ligand type, the direction of CE can
also be controlled, which enables the reverse CE process to take
place. Furthermore, we reveal the atomistic mechanism of ion
transportation in the NCs. During CE, Pb ions are kicked out by
exchanged Cd interstitials and migrate through interstitial sites.
Local structural distortions and point defects present at the PbS/
CdS interface and in the CdS shell region assist the Pb jumps by
locally expanding the voids in Cd sublattice and by reducing the
Coulombic repulsion between Pb and Cd ions.
Results
Thermodynamic driving force. As schematically shown in
Fig. 1a, CE not only involves parent NCs, but also the
surrounding solution, consisting of exchanging cations, ligands
2

and solvent molecules. CE is initiated at the interface between the

NCs and the solution and then proceeds to occur within the NCs.
Modelling the whole system atomistically requires large system
sizes and long simulation times, which is not computationally
feasible. To overcome this, we developed a coarse-grained model
to mimic the cationligand solution in which the effects of ligands
and solvent are incorporated into negatively charged large
spherical particles (Fig. 1b, the orange circles). These particles are
denoted by pseudoligands (PSLs). PbSCdS systems are modelled
using our newly developed all-atom force eld20, which
accurately describes the crystal structures, elastic properties,
polymorphic stabilities, and surface energies of PbS and CdS
solids and PbxCd1  xS mixed phases. Details about the coarsegrained PSL model and the all-atom force eld are provided in
Methods section, Supplementary Methods, Supplementary
Discussion, Supplementary Figs 1,2 and Supplementary Table 1.
The thermodynamic driving force for Pb-Cd CE is controlled
by two main factors: the free energy difference between the two
materials forming the NCs and the solubility of both cations in
the ligand solution1,2. The former depends on several factors,
such as the lattice energy, surface energy, size and shape of the
NCs, the interfacial energy when forming HNCs and the entropy
of mixing. The dominating factor is the lattice energy difference
between ZBCdS and RSPbS, as the difference in surface
energies and the lattice mismatch between ZBCdS and RSPbS
are relatively small20. During CE, the change in TDS is typically
one or two orders of magnitude smaller than that from the
enthalpy change (DH) at B500 K (ref. 24), and thus the entropy
of mixing is also not a dominating factor21. The lattice energy of

a
Anion
ligand

NC

Cations
solvent

Solution

b
Pseudoligand (PSL)

NC

Solution

Figure 1 | Schematic representations of a CE process. (a) CE in

experiments. (b) CE in MD simulations with a coarse-grained PSL model.
The dashed lines indicate the NC/solution boundaries. The black arrows
indicate the motion of the cations during CE.

NATURE COMMUNICATIONS | 7:11503 | DOI: 10.1038/ncomms11503 | www.nature.com/naturecommunications

ARTICLE

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11503

ZBCdS computed using our force eld is about 0.3 eV per

formula unit (eV/f.u.) less than that of RSPbS20. Therefore, if the
PSLCd and PSLPb interatomic interactions are equivalent (that
is, using non-preferential PSLs, see Methods section) and when
the temperature is sufciently high, the relatively low-lattice
energy of ZBCdS will promote the Pb-Cd CE process in PbS
NCs. MD simulations of a 4.7-nm-sized PbS nanocrystal
immersed in a non-preferential PSL solution were carried out
at 500 K. Pb-Cd CE was indeed observed in our simulations.
Figure 2a shows that, as a function of simulation time, the
number of Pb cations in the parent NC decreases, whereas the
number of Cd cations increases, clearly demonstrating
the exchange of cations. We also observed that the number of S
anions remained nearly constant, which indicates that, as the
anionic sublattice is more stable, the anionic sublattice is
preserved during the CE process. The CE process found in the
MD simulations is in excellent agreement with experimental
observations21: the exchanging Cd ions in the solutions initially

d
1,000

Number of ions

800
600

400
S
Pb
Cd

200
0

20

40

60

80

t (ns)

1.0
Pb-preferred PSL
Non-preferential PSL
Cd-preferred PSL

0.8

cd

0.6
0.4

cover the surface of the parent NC (that is, corresponding to the

sharp increase of the number of in-going Cd ions during the rst
few nanoseconds as shown in Fig. 2a), while the surface Pb ions in
the parent NC are dissolved into the solution (Supplementary
Movies 1 and 2). Subsequently, the Cd ions diffuse inwards into
the NCs. When the Cd ions occupy the sites of the rst few
outermost monolayers (MLs), the exchange slows down, forming
a PbS/CdS core/shell HNC (see Fig. 2e; Supplementary Movies 1
and 2). The formation of the CdS shell impedes the outward
diffusion of the Pb ions from the PbS core.
 As
the exchange slows
NC
down, the number of in-going
Cd
N
 NC 
Cd , the number of
remaining Pb in the NC NPb
and the exchange ratio (rCd),
dened as the number of in-going Cd ions divided
the
 NC by NC

NC
total number of cations in the NC (rCd NCd
= NCd
NPb ,
grey circles in Fig. 2b,c), become nearly constant after a
few tens of nanoseconds. The PbS/CdS core/shell HNCs
with higher exchange ratios (for example, the nal
conguration of an exchanged HNC at 600 K in a MD
simulation of 100 ns in Fig. 2i) have a structure and
morphology, which is very similar to those of experimentally
observed core/shell HNCs, as shown in HRTEM images of PbS/
CdS (Fig. 2j), PbSe/CdSe (Fig. 2k) and PbTe/CdTe (Fig. 2l)
congurations17,25,26.
The preferential dissolution of the cations in solutions with
different ligands has a crucial role in determining the direction
and exchange ratio of CE. Using the same system and
temperature (that is, the 4.7-nm-sized PbS NC immersed in a
CdPSL solution at 500 K), we investigated the inuence of
different ligand solutions on the CE process by considering
different PSLcation interactions. For Pb-preferred, non-preferential and Cd-preferred PSL solutions (see Methods section), the
exchange ratio rCd as a function of simulation time is shown in
Fig. 2b. As shown in Fig. 2b, Pb-Cd CE can be signicantly
accelerated while a Pb-preferred solution is used, and it can also
be completely prohibited using a Cd-preferred solution. Note that
a small amount of Cd (rCdB20%) was found in the NC with the
Cd-preferred PSLs. These Cd ions did not penetrate into the NC
but were adsorbed at the NC surface (Fig. 2f). We also found that,
by immersing a ZBCdS NC in a solution containing Pb ions and
Cd-preferred PSLs, the system underwent a reverse Cd-Pb CE

0.2
0.0

20

40

60

80

t (ns)

1.0

400 K
450 K
500 K

0.8

550 K
600 K

cd

0.6
0.4

i
0.2
0.0

20

40

60

80

t (ns)

2 nm

2 nm

Figure 2 | The Pb-Cd CE process in a 4.7-nm-sized PbS NC using

different PSLs at different temperatures. (a) Number of in-going Cd ions
(red diamonds), remaining Pb ions (blue squares) and S ions (grey circles)
in the parent PbS NCs as a function of time. MD simulations were
performed at 500 K with non-preferential PSLs. (b,c) Time evolution of
exchange ratio rCd (the ratio of in-going Cd ions over the total number of
cations in the NC). (b) Simulations performed at 500 K with different PSLs.
A constant value of rCd (B0.2) for simulation with Cd-preferred PSLs
indicates that CE did not take place. The low fraction of Cd ions
corresponds to those Cd ions adsorbed at the NC surface. (c) Simulation
performed using non-preferential PSLs at different temperatures.
(di) Snapshots of the RS-(110) sections of the NCs in 100-ns MD
simulations: (d) initial conguration; (ei) nal congurations of 100-ns
simulations at 500 K with non-preferential PSLs (e), Cd-preferred PSLs (f),
and Pb-preferred PSLs (g) and those with non-preferential PSLs at 400 K
(h) and 600 K (i). The white, blue and red spheres represent S, Pb and Cd,
respectively. (jl) HRTEM images of PbE/CdE core/shell HNCs achieved by
experimental CE, in a [110] projection. (j) A 9-nm-sized PbS/CdS core/
shell HNC, image adapted with permission from (ref. 25) (copyright 2014
American Chemical Society). (k) A 5.6-nm-sized PbSe/CdSe core/shell
HNC, image adapted with permission from (ref. 26) (copyright The Royal
Society of Chemistry 2011). (l) A 6.6-nm-sized PbTe/CdTe core/shell HNC,
image adapted with permission from (ref. 17) (copyright 2009 American
Chemical Society).

NATURE COMMUNICATIONS | 7:11503 | DOI: 10.1038/ncomms11503 | www.nature.com/naturecommunications

ARTICLE

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11503

(see Supplementary Methods and Supplementary Fig. 3). The

reverse CE did not take place in a non-preferential or a Pbpreferred PSL solution. As shown experimentally, a way to
achieve both the forward and backward Pb3Cd CE is to change
the type of ligand: Cd-oleates are often chosen as cationligands
solutions to achieve Pb-Cd CE14,19,21, whereas aqueous Pb
nitrate solutions, PbCl2oleylamine and Pboleatesoleylamine
have been used to achieve the reverse CE process2730.
The temperature is another key factor for determining the
equilibrium and the exchange rate of CE. MD simulations for PbS
nanocrystals immersed in non-preferential PSL solutions were
performed at temperatures ranging from 400 to 600 K. The CE
process was found to be faster at higher temperatures with a more
complete Pb-Cd replacement (Fig. 2c). At 600 K, nearly 90% of
the cations in the NC was replaced by Cd within 100 ns, and a
CdS shell with a thickness of about four MLs was formed
(Fig. 2c,i). In experiments, high temperatures often lead to severe
damage and distortion of the NCs (among others, Ostwald
ripening can take place)19,21. In the simulations, we also observed
that at 700 K the whole NC dissolved in the solution. Note that
temperatures used in the MD simulations should not be directly
compared with those in experiments. The timescale in our MD
simulations is generally of the order of hundreds of nanoseconds,

a
0.0020

1st outermost ML
2nd outermost ML
3rd outermost ML
4th outermost ML

 (ns13)

0.0015

0.0010

0.0005

0.0000
40

20

60
t (ns)

80

100

RMSM ()

2.0

1.5

Ion density (a.u.)

b
S
Pb
Cd

S
Pb
Cd

10

20

r ()

1.0

0.5
5

10

15

20

Self-limiting exchange. One of the appealing features of CE at

the nanoscale is the fast exchange rate compared to the process at
the macroscopic scale. Nonetheless, Pb-Cd CE is often reported
to be self-limiting. Although complete Pb-Cd exchange has
been experimentally reported16,18,21, accurate controls, high
temperatures and long times are required. Figure 3a shows the
time evolution of the averaged volume-scaled exchange rate Z (see
Methods section) at the rst four outermost MLs of the PbS NC
from the MD simulations at 550 K with non-preferential PSLs.
The CE process takes place in a layer-by-layer fashion. The
exchange front moves from the rst outermost ML to the second
ML in 20B40 ns and further proceeds to the third outermost ML
after ca. 50 ns onward (Fig. 3a; Supplementary Movies 1 and 2;
Supplementary Fig. 4). The exchange rate of the fourth layer is
always zero, suggesting that the exchange front has not yet
reached the fourth ML in 100 ns. The overall exchange rate Z
signicantly decreases as the exchange front moves inward. This
is consistent with the average root mean square motion (RMSM,
see Methods) calculated for different ionic species as a function of
the distance from the initial position of an ion to the centre of the
NC (denoted by r) as shown in Fig. 3b. To clarify the ionic
positions in the PbS/CdS core/shell structure, the radial densities
of each ionic species are plotted in the inset of Fig. 3b. The RMSM
of all types of ions increases as the ions are closer to the NC
surface, indicating a faster CE process at the surface than inside
the NC. Obviously, the large surface-to-volume ratio of
nanostructures leads to a swift CE process at the nanoscale, but
the spherical morphology of the PbS parent NCs and the layerby-layer mode of exchange in the Pb-Cd CE process determine
its self-limiting nature. This has also been reported
experimentally for an analogous Cd-Zn CE process in CdSe
NCs10.
It is instructive to point out that the radial density of each atom
type (the inset of Fig. 3b) indicates a pure PbS domain (rr15 ),
a CdxPb1xS mixed phase domain (15 oro20 ) and a pure
CdS domain (20 rro24 ). Considering the inhomogeneity in
PbS/CdS HNCs, the mixed phase layer with a thickness of 5 is
relatively thin. It has been also shown experimentally that PbS/
CdS HNCs obtained by CE have a sharp PbS/CdS boundary17,19.
Unlike systems containing materials with the same crystal
structure, which easily form mixed phases or solid solutions
(for example, the CdSe/ZnSe10 and CdSe/CdS/ZnS31 systems),
PbS and CdS in HNCs undergo a phase separation even at
relatively high temperatures24,26.

25

r ()

Figure 3 | Exchange rate and atomic mobility analysis of Pb-Cd CE.

(a) Time evolution of the averaged volume-scaled exchange rate Z (see
Methods section) for the four outermost atomic MLs. Data were collected
from 10 independent 100-ns MD simulations of a PbS NC at 550 K. The
inset shows the nal conguration of the RS-(110) section from one of the
MD simulations. The S ions are coloured according to the colour code of
the different outermost MLs. The small red and blue spheres are Cd and Pb,
respectively. (b) Averaged RMSM (see Methods section) of S (grey
circles), Pb (blue squares) and Cd (red diamonds) as a function of r (r is the
distance between the ion and the centre of the NC). The inset shows the
ion density as a function of r. The yellow region labels the CdxPb1  xS mixed
phase domain between the PbS core and the CdS shell. Data were sampled
during the last 100 ps of the ten independent 100-ns MD simulations.
4

which is much shorter than a typical timescale of the Pb-Cd CE

experiments (minshr)14,1719,21. Accordingly, a higher
temperature is required in our simulations in order to
accelerate the CE process and so that it falls within the
timescale accessible by MD.

Kinetic mechanism. Based on a variety of experimental data, two

mechanisms have been proposed for describing the kinetics of
CE: a vacancy-mediated mechanism18,21,24 and a kick-out
mechanism32. The vacancy-mediated mechanism suggests that
Pb ions from the PbS core migrate through Cd vacancies18,21,24
(Fig. 4a). These Cd vacancies form at the surface of NCs and
migrate further towards the PbE/CdE interface, transporting the
Pb ions through the CdS shell into the solution. In a recent paper,
Bothe et al.32 challenged the vacancy-mediated mechanism by
pointing out that the formation energy of Cd vacancies is rather
high, and that therefore, cations migration through vacancies is
unlikely. These authors proposed that the Pb ions are kicked out
of the PbS core by exchanged Cd ions (the kick-out mechanism)
and migrate trough interstitial sites32 (Fig. 4b). From an energetic

NATURE COMMUNICATIONS | 7:11503 | DOI: 10.1038/ncomms11503 | www.nature.com/naturecommunications

ARTICLE

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11503

Energy
Vacancy
Space

Kick

Energy

point of view, the vacancy-mediated mechanism suggests that

the Pb jumps in the CE process are assisted by energy traps
(vacancies) and thermal uctuations (Fig. 4a), whereas the kickout mechanism suggests that the Pb jumps are stimulated by
Coulombic repulsion between the Pb ions and exchanged Cd
interstitials (Fig. 4b). Due to the limitations in resolution of stateof-the-art experimental techniques, it is difcult to determine
which mechanism is correct for the Pb-Cd CE process. By
analysing the atomic trajectories in our simulations, we are able to
unveil the mechanism for CE. We divide the CE process into two
steps: (1) CE at the PbS/CdS interface and (2) the
migration in the CdS shell region. We frequently observed the
kick-out mechanism32 at the PbS/CdS interface in our
simulations. Figure 4c shows a typical kick-out motion: two
MD snapshots showing congurations before and after the jump
of a (green coloured) Pb ion out of the PbS core. An exchanged
Cd ion (the pink sphere) located at an interstitial site in the PbS
core region is present in the pre-jump conguration. The

Interstitial
site

exchanged Cd interstitial kicks the Pb ion out of the PbS core

by Coulombic repulsion and then occupies its site. We identied
many more Pb jumps at the PbS/CdS interfaces from ve
independent 200 ns MD simulations on which we performed
statistical analysis (see Supplementary Methods and
Supplementary Table 2). Exchanged Cd interstitials near the
jumping Pb ions can be found in about 47% of the identied Pb
jumps, strongly supporting the kick-out mechanism.
Another difference between the vacancy-mediated and kickout mechanisms is whether the Pb ions migrate through
vacancies or interstitial sites during CE. We examined the prejump atomistic environment (that is, the distribution of anions
and cations) around the post-jump locations of the jumping Pb
ions (see Supplementary Methods). If a Pb ion jumps into a ZB
CdS domain, it is mostly located at one of the two different cation
sites: the tetrahedral cation site (Fig. 5a,b) or the octahedral cation
site (Fig. 5c). The former is a tetrahedrally coordinated site with
four nearest neighbour anions whereby the Pb ion can be either
located at an interstitial (Fig. 5a) or at a substitutional site
(Fig. 5b). The octahedrally coordinated site has six nearest
neighbour anions so that the Pb ion can only be an interstitial
(Fig. 5c). By counting the coordination numbers of the nearest
neighbour anions around the positions in which the Pb ions
jumped, the tetrahedral and octahedral cation sites can be
distinguished (see Supplementary Methods). Figure 5d shows
the coordination numbers for the identied Pb jumps in the
interfacial and shell regions. The results indicate that at
the PbS/CdS interface, the Pb ions show a strong preference to
jump to octahedral cation sites (coordination number 6) rather
than to tetrahedral cation sites (coordination number 4). A slight

Space

c
t = 2 ps

d
Interface

0.5

Shell

Probability

0.4

0.3

0.2

0.1

Figure 4 | Vacancy-mediated mechanism versus kick-out mechanism.

(a,b) Schematic representations of the vacancy-mediated mechanism
(a) and the kick-out mechanism (b). The blue and red circles represent Pb
and Cd, respectively. The green circles denote the Pb ions migrating at the
PbS/CdS interface. The pink circle is a Cd interstitial kicking the Pb ion out of
the PbS core region. The lower panels of (a,b) schematically show the energy
landscapes of these two different mechanisms. (c) Typical snapshots from
MD simulations showing the kick-out mechanism. Left and right parts of
(c) are the congurations of the NC before and after the jump of a Pb ion
(the green sphere) out of the PbS core, respectively. The Pb jump is initiated
by a Cd interstitial (the pink sphere) in the PbS region. The blue, red and
white spheres are Pb, Cd and S, respectively. The lower panels of (c) are
magnications of the region where the kick-out motion takes place.

0.0

Coordination number

Figure 5 | Statistical analysis of different Pb jumps in the Pb-Cd CE.

(ac) Three different sites for a Pb ion in ZBCdS: (a) a tetrahedrally
coordinated interstitial site; (b) a tetrahedrally coordinated substitutional
site; (c) an octahedrally coordinated interstitial site. The blue, red and white
spheres are Pb, Cd and S, respectively. (d) Distribution of the coordination
numbers of the jumping Pb ions after the jumps at the PbS/CdS interface
and in the CdS shell region, based on an analysis of about 10,000 Pb jumps
identied from ve independent 200-ns MD simulations.

NATURE COMMUNICATIONS | 7:11503 | DOI: 10.1038/ncomms11503 | www.nature.com/naturecommunications

ARTICLE

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11503

preference for the octahedral cation sites can be found when

the Pb ions are migrating in the CdS shell region. We calculated
the averaged distances between the four nearest anions and the
jumping Pb ions ra , and those between the four nearest cations
and the jumping Pb ions rc in the post-jump congurations (see
Supplementary Table 3). The distance analysis included the Pb
jumps at the PbS/CdS interface and in the CdS shell region with
anionic coordination numbers 4 and 6. Note that coordination
numbers 3, 5 and 7 correspond to local structural distortions or
off-site Pb ions, which were excluded from the distance
calculation. The results show that the values of ra are in a
narrow range of 3.03.1 , regardless of the tetrahedral/
octahedral site type and the region (interface or shell) where
the Pb jumps took place. The values of ra are consistent with the
PbS bond length in the bulk crystal PbS (3.0 ), indicating that
the locations of the Pb ions in the CdS domains are mainly
determined by PbS interactions. For the Pb ions located at the
tetrahedral cation sites, the ratio of rc =ra is about 1.13 (see
Supplementary Table 3), which suggests that these Pb ions are
indeed interstitials (the value of rc =ra for a tetrahedrally
coordinated Pb interstitial is 1.15) rather than substitutional ions
(the value of rc =ra for a substitutional Pb is 1.63). For the Pb ions
located at the octahedral cations sites, rc =ra is about 1.08 (see
Supplementary Table 3). This value is signicantly larger than the
value of ideal Pb interstitials at octahedral cation sites (0.87). The
mismatch can only be explained by local structural distortions
and the formation of Cd vacancies. The statistical analysis of our
simulations supports the kick-out mechanism32 whereby Pb ions
migrate through interstitial sites and not through vacancies
during CE. However, the migration of the Pb ions through
octahedral sites needs to be assisted by Cd vacancies.
Discussion
The exchange rate in our simulations is faster than that in
experiments, and this Pb-Cd CE reaction rate may be attributed
to the higher reaction temperature and the higher concentrations
of Cd in the solution compared to those in experiments. In
Pb-Cd CE experiments of PbS NCs, the reaction temperature is
typically in the range of 100200 C (373473 K). In our
simulations, the highest temperature is 600 K in order to
accelerate the CE process, considering that the timescale of our
simulations is several orders of magnitude shorter than of
experiments. Such a high temperature can increase the probability
of cation jumps by two to six orders of magnitude. Raising the
temperature in the simulations bears the risk that the kinetic
mechanism found in our simulations may differ from that in
experiments at a lower temperature. However, as the atomistic
congurations of the exchanged nanocrystals in our simulations
are very similar to those observed in experiments (Fig. 2el), it is
anticipated that the mechanisms are the same. The coarse-grained
PSL model was used to mimic the ligand and solvent molecules.
The size of a PSL is smaller than that of actual ligands used in CE
experiments, and the number of the guest cations (Cd) is set equal
to the number of PSLs. Therefore, the concentration of guest
cations in the solution is higher in our simulations in comparison
to experiments (Supplementary Discussion), which may also
result in a faster exchange in the simulations. It should be noted
that our simulations are only able to show the initial stages (the
rst few hundreds of nanoseconds) of the CE process. Compared
to the exchanged NCs in experiments that has formed CdS shells
with a thickness of 12 nm (Fig. 2jl), the exchanged NC at 400 K
(within the temperature range used in CE experiments) only has
formed an incomplete and thin layer of CdS (thickness of the
shell o0.3 nm) after 100 ns (Fig. 2h), thus the CE took place to a
lesser extent. To show a CE process for this NC in a simulation at
6

400 K with a similar degree of exchange as those in experiments,

the required simulation time may be extremely long (41 s),
which is not computationally feasible.
Increasing the temperature is an effective way to accelerate the
CE process, however, it has been reported that Ostwald ripening
takes place by heating colloidal PbS NCs at 473 K, thus destroying
the monodispersity of the NCs19. A multiple step heating
scheme18,19 has been therefore used for CE experiments at high
temperatures in which the NCs are protected by CdS shells that
form at a lower temperature. We did nd signicant dissolution
of S ions in our simulations at high temperatures (Supplementary
Movies 1 and 2; Supplementary Fig. 5), however, Ostwald
ripening could not be directly observed in our simulations since a
single NC was considered. The dissolved S anions may also
recrystallize on other parts of the same NC with the cations in the
solution, thus enabling massive migration of anions and
reconstruction of PbE/CdE HNCs. This phenomenon becomes
more important at high temperatures and/or with active ligands,
in some extreme cases PbE NCs or PbE/CdE core/shell HNCs
may transform to a Janus-like heterostructure26,29,33.
Besides the multiple step heating scheme, we found that
increasing the pressure could be an alternative approach to
preserve NCs at elevated temperatures in order to accelerate the
CE processes and to achieve complete compositional conversions.
The Pb-Cd CE process cannot be accelerated by merely
applying a hydrostatic pressure to the solution (Supplementary
Methods and Supplementary Fig. 6), however, an increased
hydrostatic pressure is able to increase the melting point of ionic
semiconductor solids34 and their stability at high temperatures.
For example, only ve Pb ions remained in the exchanged NC
(rCd 0.997) in a MD simulation of 100 ns at 750 K and a
hydrostatic pressure of 1.00 GPa, and a PbS NC completely
transformed into a CdS NC in 70 ns at 800 K and 1.0 GPa
(Supplementary Fig. 6). At a standard pressure the NC would
have been severely damaged upon reaching a temperature of
700 K, whereas the exchanged NC remains solid phase and retains
a quasi-spherical morphology at high-temperaturehigh-pressure
conditions (p 1.00 GPa, Tr800 K, Supplementary Fig. 6). Our
simulations therefore suggest that a high-pressurehightemperature approach is a promising method to control the
rate and degree of the Pb-Cd CE process.
The exchange process is initiated by Cd interstitials in the PbS
domain, whereby these Cd interstitials kick Pb ions out of the PbS
core. A high temperature is needed to promote the penetration of
these guest cations into the NC, and this explains why some types
of CE processes (for example, Pb-Cd and Zn-Cd) only take
place upon annealing. Although it has been shown that Pb ions in
Pb-Cd CE migrate through interstitial sites, it is important to
note that Cd vacancies, or most likely, Cd Frenkel pairs (a Cd ion
moves to a neighbouring interstitial site, thereby forming a
vacancy and an interstitial) play a role during the exchange
process, especially at the PbS/CdS interface. Since PbS has a
sixfold RS structure, there should be a strong preference for Pb
ions to occupy octahedral cation sites. However, in CdS domains,
Pb ions are prevented from jumping into octahedral interstitial
sites due to strong Coulombic repulsion between the Pb ions and
the densely packed Cd ions around the octahedral interstitial
sites. Only when assisted by Cd vacancies, which are present due
to the formation of Cd Frenkel pairs, Pb ions can jump into
octahedral cation sites. It is well known that structural instability
will be introduced by dipole moments at the polar/non-polar
interface or the polar/polar interface with different crystal
structures35. Therefore, formation of Cd Frenkel pairs and local
structural distortion occurs more easily and more frequently at
the PbS/CdS interface than in the CdS domain at high
temperatures. This conclusion agrees with the results from our

NATURE COMMUNICATIONS | 7:11503 | DOI: 10.1038/ncomms11503 | www.nature.com/naturecommunications

ARTICLE

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11503

MD simulations showing that the Pb ions show a stronger

preference for octahedral cation sites while jumping at the PbS/
CdS interface. In general, the CE processes take place more easily
and faster in the systems forming interfaces of two different
structures (for example, PbE/CdE14,17,18, Ag2E/CdE22 and Cu2E/
CdE36) than those having the same structure (for example, CdE/
ZnE10,31). In addition, we point out here that the vacancymediated mechanism wrongly guessed the detailed kinetics of
this particular CE process. Cd vacancies do not migrate back and
forth between the NC surface and the PbS/CdS interface, and Pb
ions do not migrate through Cd vacancies. The formation of Cd
vacancies (without Cd Frenkel pairs) is indeed difcult32,
especially at high temperatures whereby vacancies tends to be
annealed out36. Pb substitutions in the CdS shell region have been
rarely observed in our simulations.
As CdS and PbS have different crystal structures, a structural
transition must take place during a CE process. Two most stable
phases of CdS, wurtzite (WZ) and ZB, have almost the same
cohesive energy (the difference between the cohesive energies of
these two CdS polymorphs is about 1 meV/f.u. according to
density functional theory (DFT) calculations20). During the CE
process of Pb and Cd chalcogenides nanostructures, however, the
RS-to-ZB transition has been more frequently observed
experimentally than the RS-to-WZ transition1618,23,24 (see
Fig. 2jl). The latter requires an extra step of anionic sublattice
slips37 as the former only requires slips of the cationic
sublattice20. Therefore, the RS-to-ZB transition has a lower
energy barrier than the RS-to-WZ transition. In our simulations,
almost equal amounts of the WZ and ZB domains were found in
the CdS shell of an exchanged HNC (Supplementary Fig. 7). This
discrepancy between experiments and simulations may be
attributed to the adopted force eld for CdS as the stability of
ZB phase is slightly underestimated20. However, we note that the
competition between the RS-to-ZB and RS-to-WZ transitions in
Pb-Cd CE also depends on many other factors, including the
morphology of initial parent NCs, types of ligand and solvent,
temperatures and nucleation energies. For instance, the WZ
stacking faults and even direct RS-to-WZ transition have been
observed in CE experiments1618,23,24. In our simulations, most
of the Cd ions were located at the tetrahedral cation sites (ZB or
WZ), and only those at or near the PbS/CdS interfaces may jump
between tetrahedral and octahedral sites. It has been reported that
CdS shells may remain in a metastable RS phase during CE
processes25, which has not been observed in our simulations.
In summary, we used classical MD simulations to investigate
Pb-Cd CE in colloidal PbS NCs. The cation-ligands solutions
were simulated using a coarse-grained PSL model, allowing us to
study this large system at a long-time scale. The main driving
force of the Pb-Cd CE process is the competition between the
ionic crystal lattice energies and the cations solvent solubility. By
adjusting the PSLcation interatomic interactions (analogous to
changing the cationligand solutions in experiments) and the
temperature, the direction and rate of the CE process can be
controlled. Our simulation results reveal that the CE process is
initiated by kick-out motions and that Pb ions migrate through
interstitial sites. Highly mobile Cd ions at the PbS/CdS interface
facilitate the formation of Cd Frenkel pairs, which enables the
jumps of cations and accelerates the CE process. Our study
provides a profound insight into the physics underlying the
Pb-Cd CE at the atomic level. The hybrid all-atom/coarsegraining approach is able to effectively reveal both the thermodynamics and kinetics of the Pb-Cd CE, and, more importantly,
this model may also be adapted to investigate other types of CE
(for example, Cd-Cu and Zn-Ag) and other synthesis
techniques (for example, hot injection, seeded growth and
vapourliquidsolid growth). Finally, our simulations also predict

that high-temperaturehigh-pressure conditions can greatly

enhance the CE rate. To date, several key questions remain
unclear, such as: Why does CE behave very differently in different
materials? What is the optimal ligand solution, temperature and
pressure to yield complete Pb-Cd CE? How can the thickness
and uniformity of the CdE shell be completely controlled? Our
approach provides opportunities to further address these
questions, and we anticipate that the coarse-grained PSL model
will be extended and further improved.
Methods
Colloidal PbS NC systems and MD simulations. A spherical PbS NC with a
diameter of 4.7 nm containing 1,088 {PbS} pairs was cut from a RSPbS matrix
with a lattice parameter of 6.0 . The PbS NC was placed in the centre of a cubic
simulation box with a size of 10 nm. The simulation box was then randomly lled
with 4,352 {Cd-PSL} pairs (to avoid overlaps, we ensured that the distance between
any two ions is larger than 3.5 ). Periodic boundary conditions were applied in all
directions.
In MD simulations, the PSLPSL or PSLcation interatomic interactions consist
of two parts: long-ranged Coulombic interactions and short-ranged interactions
described by a Buckingham potential. We assume that electron transfer does not
play an important role during Pb-Cd CE, that is, the charges of the cations,
anions and anionic ligands in the solid and liquid phases, and at the solid/liquid
interface, do not change. The main function of our PSL model is to create a
solution in which the guest cations are dissociated and CE takes place at the
NC/solution interface (see Supplementary Discussion). The parameters in the
Buckingham potentials can be altered to create Pb-preferred, Cd-preferred or
non-preferential PSLs (Pb/Cd-preferred means that the Pb/CdPSL interactions
are stronger than the Cd/PbPSL interactions; non-preferential means that the
CdPSL and PbPSL interactions are equivalent, see Supplementary Discussion,
Supplementary Table 1 and Supplementary Fig. 1), so that the inuence of the
cation-ligand interactions on CE can be quantitatively investigated.
All MD simulations were carried out using the LAMMPS code38
(http://lammps.sandia.gov). The particleparticleparticlemesh method39 was
used to calculate the Coulombic interactions and a cut-off radius of 10 was set for
all short-ranged interactions. The equations of motion were integrated using the
velocityVerlet algorithm with a time step of 1 fs. Four S ions in the centre of the
NC were constrained at xed positions to prevent net motion and rotation of the
NCs. Simulations of a few hundreds ns were carried out in the NPT ensemble and
the rst 100 ps was used for preliminary equilibration at the desired temperatures
and zero pressure. The longest simulation time was 300 ns.
Volume-scaled exchange rate. Several MLs were dened in a PbS NC (one ML
has a thickness of 3.5 , see the inset of Fig. 3a). The exchange rate in the jth ML
and at time t, Zj(t), was dened as:
 o
  o

i
i
N  NCd
 NPb  NPb
Zj t Cd
2
DtDVj
where N is the number of Cd or Pb ions passing through the surface o or i during a
time Dt. The superscripts o and i represent the outer and the inner surfaces of the
jth ML, respectively. A positive value of N indicates the cations are moving inward
to a given ML, while a negative value indicates that the cations are moving
outward. Dt is a small time interval (Dt 1 ps) and DVj is the volume of the jth ML.
Using this denition, a positive value of Z indicates a Pb-Cd CE process and a
negative value indicates the reverse Cd-Pb CE. We sampled the Zj every 100 ps
for the rst four outermost MLs from simulations for a PbS NC with non-preferential PSLs at 550 K. Each data point in Fig. 3a corresponds to an averaged Zj
over a time interval of 10 ns from 10 independent MD simulations.
Root mean square motion. The RMSM was used to describe the vibration and
mobility of ions. The RMSM was dened as:
v
u X
u1 t   
2
3
RMSM t
xi tj  xi
t tj 1
where t is the time to sample the data, which is the last 100 ps of 10 independent
100- ns simulations. xi(tj) is the position of ion i at time tj, 
xi is the time-averaged
position of the ion i. The time interval between two samplings is 1 ps. The RMSM
of each ion and its initial distance r to the centre of the NC were calculated. The
RMSM was further averaged over the same types of ions with similar values of
r (0.5 ).

References
1. Beberwyck, B. J., Surendranath, Y. & Alivisatos, A. P. Cation exchange: a versatile
tool for nanomaterials synthesis. J. Phys. Chem. C 117, 1975919770 (2013).

NATURE COMMUNICATIONS | 7:11503 | DOI: 10.1038/ncomms11503 | www.nature.com/naturecommunications

ARTICLE

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11503

2. Rivest, J. B. & Jain, P. K. Cation exchange on the nanoscale: an emerging

technique for new material synthesis, device fabrication, and chemical sensing.
Chem. Soc. Rev. 42, 8996 (2013).
3. Gupta, S., Kershaw, S. V. & Rogach, A. L. 25th anniversary article: ion exchange
in colloidal nanocrystals. Adv. Mater. 25, 69236944 (2013).
4. Son, D. H., Hughes, S. M., Yin, Y. & Alivisatos, A. P. Cation exchange reactions
in ionic nanocrystals. Science 306, 10091012 (2004).
5. Robinson, R. D. et al. Spontaneous superlattice formation in nanorods through
partial cation exchange. Science 317, 355358 (2007).
6. Li, H. et al. Sequential cation exchange in nanocrystals: preservation of crystal
phase and formation of metastable phases. Nano Lett. 11, 49644970 (2011).
7. Zhang, B., Jung, Y., Chung, H.-S., Vugt, L. V. & Agarwal, R. Nanowire
transformation by size-dependent cation exchange reactions. Nano Lett. 10,
149155 (2009).
8. Miszta, K., Dorfs, D., Genovese, A., Kim, M. R. & Manna, L. Cation exchange
reactions in colloidal branched nanocrystals. ACS Nano 5, 71767183 (2011).
9. Jain, P. K., Amirav, L., Aloni, S. & Alivisatos, A. P. Nanoheterostructure cation
exchange: anionic framework conservation. J. Am. Chem. Soc. 132, 99979999
(2010).
10. Groeneveld, E. et al. Tailoring ZnSeCdSe colloidal quantum dots via cation
exchange: from core/shell to alloy nanocrystals. ACS Nano 7, 79137930
(2013).
11. Wark, S. E., Hsia, C.-H. & Son, D. H. Effects of ion solvation and volume
change of reaction on the equilibrium and morphology in cation-exchange
reaction of nanocrystals. J. Am. Chem. Soc. 130, 95509555 (2008).
12. Murray, C. B., Norris, D. J. & Bawendi, M. G. Synthesis and characterization of
nearly monodisperse CdE (E sulfur, selenium, tellurium) semiconductor
nanocrystallites. J. Am. Chem. Soc. 115, 87068715 (1993).
13. Sadtler, B. et al. Selective facet reactivity during cation exchange in cadmium
sulde nanorods. J. Am. Chem. Soc. 131, 52855293 (2009).
14. Pietryga, J. M. et al. Utilizing the lability of lead selenide to produce
heterostructured nanocrystals with bright, stable infrared emission. J. Am.
Chem. Soc. 130, 48794885 (2008).
15. Jia, G. et al. Couples of colloidal semiconductor nanorods formed by selflimited assembly. Nat. Mater. 13, 301307 (2014).
16. Boneschanscher, M. P. et al. Long-range orientation and atomic attachment of
nanocrystals in 2D honeycomb superlattices. Science 344, 13771380 (2014).
17. Lambert, K., Geyter, B. D., Moreels, I. & Hens, Z. PbTe|CdTe core|shell particles
by cation exchange, a HR-TEM study. Chem. Mater. 21, 778780 (2009).
18. Casavola, M. et al. Anisotropic cation exchange in PbSe/CdSe core/shell
nanocrystals of different geometry. Chem. Mater. 24, 294302 (2011).
19. Zhao, H., Chaker, M., Wu, N. & Ma, D. Towards controlled synthesis and
better understanding of highly luminescent PbS/CdS core/shell quantum dots.
J. Mater. Chem. 21, 88988904 (2011).
20. Fan, Z. et al. A transferable force eld for CdSCdSePbSPbSe solid systems.
J. Chem. Phys. 141, 244503 (2014).
21. Justo, Y. et al. More cation exchange and the chemistry of the nanocrystal
surface. ACS Nano 8, 79487957 (2014).
22. Chan, E. M. et al. Millisecond kinetics of nanocrystal cation exchange using
microuidic X-ray absorption spectroscopy. J. Phys. Chem. A 111,
1221012215 (2007).
23. Bals, S. et al. Three-dimensional atomic imaging of colloidal coreshell
nanocrystals. Nano Lett. 11, 34203424 (2011).
24. Yalcin, A. O. et al. Atomic resolution monitoring of cation exchange in
CdSe-PbSe heteronanocrystals during epitaxial solidsolidvapor growth. Nano
Lett. 14, 36613667 (2014).
25. Lechner, R. T. et al. Crystal phase transitions in the shell of PbS/CdS core/shell
nanocrystals inuences photoluminescence intensity. Chem. Mater. 26,
59145922 (2014).
26. Grodzinska, D., Pietra, F., van Huis, M. A., Vanmaekelbergh, D. &
de Mello Donega, C. Thermally induced atomic reconstruction of PbSe/CdSe
core/shell quantum dots into PbSe/CdSe Bi-hemisphere hetero-nanocrystals. J.
Mater. Chem. 21, 1155611565 (2011).
27. Zhou, H. S., Honma, I., Komiyama, H. & Haus, J. W. Coated semiconductor
nanoparticles; the CdS/PbS systems synthesis and properties. J. Phys. Chem. 97,
895901 (1993).

28. Zhang, J. et al. PbSe quantum dot solar cells with more than 6% efciency
fabricated in ambient atmosphere. Nano Lett. 14, 60106015 (2014).
29. Zhang, J. et al. Preparation of Cd/Pb chalcogenide heterostructured janus
particles via controllable cation exchange. ACS Nano 9, 71517163 (2015).
30. Lee, D. et al. Direct Cd-to-Pb exchange of CdSe nanorods into PbSe/CdSe axial
heterojunction nanorods. Chem. Mater. 27, 52955304 (2015).
31. Yalcin, A. O. et al. Heat-induced transformation of CdSeCdSZnS Core
multishell quantum dots by Zn diffusion into inner layers. Chem. Commun. 51,
33203323 (2015).
32. Bothe, C. et al. Solid-state chemistry on the nanoscale: ion transport through
interstitial sites or vacancies? Angew. Chem. Int. Edit 54, 1418314186 (2015).
33. Habinshuti, J. et al. Phase transformation of PbSe/CdSe nanocrystals from
coreshell to janus structure studied by photoemission spectroscopy. Phys. Rev.
B 87, 184102 (2013).
34. Li, D., Zhang, P. & Yan, J. Ab initio molecular dynamics study of high-pressure
melting of beryllium oxide. Sci. Rep. 4, 4707 (2014).
35. Harrison, W. A., Kraut, E. A., Waldrop, J. R. & Grant, R. W. Polar
heterojunction interfaces. Phys. Rev. B 18, 4402 (1978).
36. Lesnyak, V., Brescia, R., Messina, G. C. & Manna, L. Cu vacancies boost cation
exchange reactions in copper selenide nanocrystals. J. Am. Chem. Soc. 137,
93159323 (2015).
37. Fan, Z. et al. From sphere to multipod: thermally induced transitions of CdSe
nanocrystals studied by molecular dynamics simulations. J. Am. Chem. Soc.
135, 58695876 (2013).
38. Plimpton, S. Fast parallel algorithms for short-range molecular dynamics.
J. Comput. Phys. 117, 119 (1995).
39. Darden, T., York, D. & Pedersen, L. Particle mesh Ewald: an Nlog(N) method
for Ewald sums in large systems. J. Chem. Phys. 98, 1008910092 (1993).

Acknowledgements
Z.F. acknowledges Dengpan Dong from University of Utah for useful discussions of the
kinetic mechanism. This work is part of the research program of the Foundation for
Fundamental Research on Matter (FOM), which is part of the Netherlands Organisation for
Scientic Research (NWO). This work was sponsored by NWO Exacte Wetenschappen
(Physical Sciences) for the use of supercomputer facilities, with nancial support from the
Nederlandse Organisatie voor Wetenschappelijk Onderzoek (Netherlands Organisation for
Scientic Research, NWO). M.A.v.H. acknowledges NWO for a VIDI grant. The data that
support the ndings of this study will be stored in a repository, and are available from the
authors upon request.

Author contributions
Z.F., T.J.H.V. and M.A.v.H. conceived the study. Z.F. developed the simulation model
and performed the simulations. Z.F., L.-C.L. and W.B. performed the analyses. All
authors discussed the results and participated in writing the manuscript.

Additional information
Supplementary Information accompanies this paper at http://www.nature.com/
naturecommunications
Competing nancial interests: The authors declare no competing nancial interests.
Reprints and permission information is available online at http://npg.nature.com/
reprintsandpermissions/
How to cite this article: Fan, Z. et al. Atomistic understanding of cation exchange in PbS
nanocrystals using simulations with pseudoligands. Nat. Commun. 7:11503 doi: 10.1038/
ncomms11503 (2016).
This work is licensed under a Creative Commons Attribution 4.0
International License. The images or other third party material in this
article are included in the articles Creative Commons license, unless indicated otherwise
in the credit line; if the material is not included under the Creative Commons license,
users will need to obtain permission from the license holder to reproduce the material.
To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

NATURE COMMUNICATIONS | 7:11503 | DOI: 10.1038/ncomms11503 | www.nature.com/naturecommunications